dƒusion - fair & decentralized exchange

​

Research�Area

Plasma

Snapp

Plasma

Observation & Predictions

Market Mechanism

Scalable Batch Auctions

​

Today

Future

Trading volume is focused on centralized exchanges

Trading volume will shift to

decentralized exchanges �

​

Today

Future

Centralized matching engine,

front-running

Decentralized matching engine,

no front-running

​

​

​

SOLVER1

SOLVER2

price1

price2

SOLVER3

price3

​

Today

Future

Many Stable Coins �will change the picture

Dominant trading pairs

Observation & Predictions

Market Mechanism

Scalable Batch Auctions

​

Batch auctions for 2 tokens: [A,B]

​

Orders �A -> B

​

Orders �B -> A

EXCHANGE RATE A VS. B

TRADING VOLUME [A]

OPTIMAL PRICE

Collect orders over a predefined time

Generation of order book

Calculate the optimal price maximizing traders surplus

Settle all orders with the optimal price

Traders�surplus�

Ring trades

​

SELL A FOR B

---

SELL B FOR C

​

SELL C FOR A

Ring

Trades

Multi-dimensional order-books

​

PRICE A/B

Traders Surplus

PRICE B/C

PRICE C/A

TRADING 3 TOKENS

OPTIMAL UNIFORM �CLEARING PRICE

Executing all ring-trades at the best possible price

Multi-token batch auction mechanism

p(B|A)

p(C|B)

Order can trade any token against any token

Order collection for x minutes

Computation of uniform clearing prices, i.e.

• Maximum traders surplus
• Arbitrage-freeness:

p(C|B) * p(B|A) = p(C|A)

• Constraints:

Sum of tokens X sold equals sum � of tokens X bought (for all tokens)

​

Settlement of orders

​

Today

Future

Many Stable Coins �will change the picture

Dominant trading pairs

Token growth thesis

�200 tokens created by day

Trading Volume

Popular regulated Tokens

Low liquidity tokens�(Prediction tokens, �Uniswap tokens,�CPDs, tokenized assets of games)��

Popularity of Token

Dfusion is a decentralized trading protocol, which allows anyone to list/trade even unpopular tokens. �

Target

Observation & Predictions

Market Mechanism

Scalable Batch Auctions

1. Solution:��Batch auctions

on plasma

Plasma

Expectation for the plasma exchange:

• 3 min batch time
• 60k orders capacity per batch
• 150 tps capacity built on plasma MVP
• Per trade gas costs are reduced significantly

​

Today

Future

Trading volume is focused on centralized exchanges

due to scalability

Trading volume will shift to

decentralized exchanges �

Plasma ...

… Snapp

• Scales well
• Low operation cost�
• complicated exit games
• data availability not guaranteed
• some degree of centralization (MVP)

On-chain scaling via snapp�Snark application (snapp)

​

Complete snapp states are represented via root hashes

Transaction are collected on chain as payload and hashed to H(t)

Anyone can calculate the snark for the state transition �(r1 ,H(t))→ r2

Same security as ETH

�25x scaling of transactions

​

Each leaf node represents a data in the snapp

r1

Tx

Tx

Tx

Snark:�(r1 ,H(t))→r2

r2

H(t)

Deposits and exits in snapps

​

Users need to request the exits/ deposits on-chain via

Information about exits and deposits queues are stored in hashes H(q)

​

Anyone can calculate the snark processing the queue-hashes, proving the state transition�(r1 ,H(q))→ r2�

Each leaf node represents a state in the snapp

r1

deposit

exit

deposit

Snark:�(r1 ,H(q))→r2

r2

H(q)

deposit

Plasma ...

… Snapp

• Scales well
• Low operation cost�
• complicated exit games
• data availability not guaranteed
• some degree of centralization (MVP)

​

• Data availability solved
• Scales reasonable
• Enables to build more advanced features�
• Operation costs higher
• Expensive snarks��Later on these solution can be ported to Plasma (snapp)

2. Solution:��Batch auctions

as snapp

Snapp

Expectation for the dfusion:

• Quick batch times
• 1K orders + execution costs only 7.4M gas
• Fully decentralized trading engine

​

Dfusion

dƒusion - orders collection

Traders send signed orders to an operator or different operators

Operators bundle orders and send it on-chain

​

On-chain orders are getting hashed together with into H(o)

orders

orders

order

H(o)

order

order

order

dƒusion - price calculation

Once orders are final, everyone can try to find the best uniform clearing price

Every solver can submit his solution within x minutes

​

The smart contract chooses the price as final price, which maximizes the traders surplus

orders

orders

order

Max( surplus(p) )

H(p)

The solver needs to provide a bond and will be slashed if the solution is not correct.

SOLVER1

SOLVER2

price1

price2

SOLVER3

price3

​

Today

Future

Centralized matching engine,

front-running

Decentralized matching engine,

no front-running

​

​

​

SOLVER1

SOLVER2

price1

price2

SOLVER3

price3

dƒusion - settlement

Best prices are determined by the smart contract

Submitter of best solution must calculate new state rn+1 and the snark proving the state transition from

(rn , H(o), p) →rn+1

with the orders encode in H(o) and the prices p

​

Max([surplus])

Snark:�(rn ,H(o),p)→rn+1

rn

rn+1

dƒusion - Process picture

​

orders

orders

order

H(o)

order

order

rn

rn+1

Max( surplus(p) )

price1

price2

Snark:�(rn ,H(o),p)→rn+1

Order Collection

Price calculation

Settlement

dƒusion - Process picture

​

orders

orders

order

H(o)

order

order

rn

rn+1

Max( surplus(p) )

price1

price2

Resolution via Snark:�(rn ,H(o),p)→rn+1

Order Collection

Price calculation

Settlement

Bonded Proposal:�(rn ,H(o),p)→rn+1

Bonded Challenge

Snark:�(rn ,H(o),p)→rn+1

Take away picture

​

PRICE A/B

TRADING SURPLUS

PRICE B/C

PRICE C/A

TRADING 3 TOKENS

OPTIMAL UNIFORM �CLEARING PRICE

Executing all ring-trades at the best possible price

Thank you for your attention!

And special thanks to the team:

Tom Walther

Ben Smith

Stefan George

Martin Köppelmann

Alexey Akhunov

Felix Leupold

Christian Reitwiessner

Johann Barbie

Alan Lu

Collin Chin

Dominik Teiml

Julian Garcia